Reflection probe array can exceed the driver's per-allocation limit (maxMemoryAllocationSize) #1081
Labels
No labels
amd
avatars
bug
builder
cannot-reproduce
crash
epic
feature
from-viewer
gl-removal
known-limitation
monitoring
parity
perf
platform
priority
critical
priority
high
priority
low
priority
medium
provenance
rendering
rt
status
in-progress
status
testing
status
up-next
triage
ui
upstream
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
sky_line/Slipstream#1081
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Found by the automated validation run on the 1.5 candidate (GTX 1080): setting reflection probes to 256 slots at 512 resolution allocates the probe array as one 4.30 GB image; the device's
maxMemoryAllocationSizeis 4.29 GB.The VRAM check (
vramPick, called fromconfigureProbes) only compares against free VRAM, so it reported "fits". NVIDIA accepts it anyway; drivers that enforce the limit refuse it and the soft allocation falls back to a smaller set (no crash), but it is still out of spec.Fix: query
VkPhysicalDeviceMaintenance3Properties::maxMemoryAllocationSizeonce at device init and makevramPick(or the probe candidate list) skip any candidate whose largest single allocation exceeds it. Check other large single allocations (shadow atlases, GI/ReSTIR buffers, the vis queue) against the same limit.